1. Drift and Behavior of E. coli Cells
Gabriele Micali, Rémy Colin, Victor Sourjik, Robert G. Endres 
Biophysical Journal 
Volume 113, Issue 11, Pages (December 2017)
DOI: /j.bpj
Copyright © 2017 Biophysical Society Terms and Conditions

2. Figure 1
Schematic of the experimental setup. A chemical gradient of MeAsp (a nonmetabolizable analog of the amino acid Asp) is created in a microfluidic device by maintaining a fixed concentration on one side of the channel and zero on the other. E. coli cells (strain MG1655) are injected on both sides and free to move in aerobic conditions. The gradient is stable after ∼1 h 30 min and the data were acquired after 2, 3, and 4 h with each experiment repeated three times (see the Supporting Material). The gradient was measured after the final acquisition using fluorescein. (Middle) Fluorescence picture of the microfluidic chamber is given, with the white bar representing 500 μm. (Left and right) Exemplar given of single E. coli trajectories from a typical movie, acquired in the middle of the channel with trajectory starting points marked with black dots (dashed box; for details, see the Supporting Material). Some of the trajectories are relatively straight (left), whereas others are curled (right). The MeAsp gradient is oriented to the right in this image (lighter shading corresponds to higher ligand concentration). The average concentration in the channel was 1 mM. To see this figure in color, go online.
Biophysical Journal  , DOI: ( /j.bpj )
Copyright © 2017 Biophysical Society Terms and Conditions

3. Figure 2
Swimming behavior of E. coli cells from simulations. (a) Trajectories of cells chemotacting in an exponential gradient of MeAsp from our simulations based on a modified version of the software RapidCell (see the Supporting Material for details) (15). Cells are initially placed at x0 = 2.5 mm and y0 = 10 mm (note that the y position is irrelevant, as the gradient is independent of y). We impose reflecting boundary conditions at x = 0, y = 0, and y = 20 mm, and an adsorbing boundary condition at x = 30 mm. (Inset) Trajectories given without rotational diffusion. Relative gradient is λ = 0.75 mm−1. (b) Given here is the average drift velocity from simulations (solid lines) and analytical (dashed lines) theory. Vertical and horizontal dotted lines indicate the discrepancy between theory and simulations due to finite number of methylation sites (vertical) and very steep gradients (horizontal). Note that this averaged single-cell drift from trajectories is equivalent to the steady-state drift from experiments (next figure). For both panels, parameters are reported in Table S1. To see this figure in color, go online.
Biophysical Journal  , DOI: ( /j.bpj )
Copyright © 2017 Biophysical Society Terms and Conditions

4. Figure 3
Experimental verification of peak using microfluidics. Comparison of drift from theory, simulations, and experiments. Given here are experimental data in linear gradients (dots); each data point is the average over nine measurements (three measurements at different times in triplicates). Vertical bars are error of the mean, whereas horizontal bars range from low to high concentrations observed during experiments (relative gradients are ∼0.45 mm−1, cells swim with an average velocity of ∼11.8 ± 1.3 μm s−1; see the Supporting Material). Horizontal bar indicates osmotic stress regime (25). Theory (smooth lines) and simulations (jagged lines) in exponential gradient (λ = 0.45 mm−1) with upper (13.1 μm s−1) and lower (10.5 μm s−1) velocity bounds as shown (2). Parameters for theory and simulations are reported in Table S1. (Insets) Shown is a comparison of distributions of drift velocities (μm/s) from unfiltered and filtered experimental trajectories, where the filter selected trajectories with an average run velocity of ± 1.31 μm/s, as well as trajectories from simulations for concentrations at ∼0.1 mM (top). The bimodal distribution from simulations is due to projection along the x axis, whereas the center peak of the unfiltered data is due to tumbles of finite duration, with the peak height reduced when filtered for speed. (Bottom) Shown here is the piecewise linear approximation of the exponential gradient used in the experiments. To see this figure in color, go online.
Biophysical Journal  , DOI: ( /j.bpj )
Copyright © 2017 Biophysical Society Terms and Conditions